Skip to content

Research: CK dyadic ladders save 13,770 / 67,146 primitives (no new kappa) - #769

Merged
Mister-Meeseeks merged 1 commit into
CrocSwap:mainfrom
24kHandsome1201:research/complex-dyadic-runs-seed36
Oct 14, 2026
Merged

Mister-Meeseeks merged 1 commit into
CrocSwap:mainfrom
24kHandsome1201:research/complex-dyadic-runs-seed36

Conversation

@24kHandsome1201

Copy link
Copy Markdown
Contributor

The selected #626 CK programs prepare consecutive dyadic multiples of an unchanged source from scratch. This finite circuit contribution retains coefficient work across 743 varying-magnitude runs / 10,124 additions, using exact halving/doubling when cheaper than resetting.

Status: finite complex-supplier circuit improvement / research note; no new kappa.

CK scalar program Selected real primitives Replacement Saved
Forward 224,938 211,168 13,770 (6.12%)
Backward / inverse-transpose 224,986 157,840 67,146 (29.84%)

All 463 forward runs are in full-frame cleanup; all 280 backward runs are in zero-frame preconditioning. Source reads, initialization, shifts, signed target updates and clearing are fully charged. Exact positive prefix envelopes remain unchanged, as do the macro word, target-update order, native frames, helpers and paid child profile. These are circuit counts, not runtime measurements.

Added evidence and concurrent credit

Credit @ellipsi2's #722 for equal-magnitude sharing. On these same CK runs, changing magnitudes adds 6,012 forward + 12,596 backward primitives saved beyond an independently recounted equal-only comparator.

Concurrent #739 by @namredips published the same scale-reuse mechanism for JP8 forward cleanup while this package was being verified. This submission claims no priority for that mechanism. It adds separately implemented CK coverage in both orientations, particularly the backward preconditioning phase, and fresh complete boundary/prefix evidence. The final scan through #767 also found @taizaki69's #755 (a1dd9f819102c40e0b5e4a75215cf69ee864f480), which covers equal-magnitude JP8 reuse in both orientations with #696's copy initialization. That backward-phase extension is credited; no phase-priority claim is made. This package covers CK varying magnitudes under original Horner initialization. No pending patch is imported. #696/#723/#734 single-macro lowering, #750 live-scale relocation and #751 JP8 common-target fan-in are distinct and not combined with these counts.

Only research/complex-ck-dyadic-ladders/ is added: executable circuits, source pins, complete per-run receipt, proof, two verification implementations, negative controls, attribution and local claim.

Reproduce

python3 -I -B research/complex-ck-dyadic-ladders/verify.py
python3 -I -B research/complex-ck-dyadic-ladders/check_columns.py

Expected final lines:

PASS_CK_DYADIC_LADDERS blocks=743 saved=13770,67146 columns=716+9 controls=13
PASS_CK_LADDER_COLUMNS blocks=743 columns=11610 missing_clear_rejections=743

Linux CPython 3.12.14, standard library only. Both complete scalar projections check all 716 live + nine arbitrary private input columns. A separate flat-vector implementation checks 11,610 local basis cases, with missing-final-clear rejection in all 743 blocks. Thirteen main controls include endpoint-neutral prefix inflation, sign/order errors, source writes, frame/phase crossing and undercharging. Isolated optimized replay, minimal relocated checkout, alternate gzip packaging, source/receipt corruption controls, output-preservation, selected-record and staged whitespace checks pass.

Source and scope

Base: 66849e1475c930fd7732b1a0067ffee7e1e95f60. Selected #626: 660de893d27bb7abd1d5b01b1640207528a5c851; reviewed #622: f69477f4f6d3c7eb1f740b166881bffed4ffa9f6. OS assignment seed 36, complex track, CK selected by parity. No pinned coordination issue existed, so the claim/result are local.

The private-work clearing contract moves to fused-block boundaries. Production-emitter integration and refreshed global invoice/admission remain separate obligations. Full repository verification, cold supplier regeneration, native phase/wrapper regeneration, global admission, Lean and remote CI were not run. All inherited compiler, common-label/mutable-basis realization, hosting, precision/tape/recovery, bit and analytic assumptions remain. The selected exponent is unchanged.

Credit Jeremy Kothe and the #622 contributors, Chafik Boukhalfa's #626 packaging, icekylinx's factory/emitter lineage, DreamingOfClouds, Dugongue, eumemic, Rohan Arun and retained predecessors. Prepared for @24kHandsome1201 with substantial OpenAI Codex assistance. Both verification implementations were written by one AI agent; no independent-human or delegated-agent review is claimed. Apache-2.0.

@Mister-Meeseeks
Mister-Meeseeks merged commit 0dc28e2 into CrocSwap:main Oct 14, 2026
@Mister-Meeseeks

Copy link
Copy Markdown
Contributor

100-PR batch review at e891b2b19e3185ec91bcf001e255af87f42389ce.

Merged to main, preserving the original contributor commits and package attribution.

Reviewed scope: CK dyadic ladders: complete coefficient reuse, arbitrary logical/private columns and unchanged recursive events; local count savings only.

The pinned scoped command passed; the batch ledger contains its command, input hashes, environment and log. This incorporates the research/tooling contribution and does not change the selected κ or discharge the inherited multiplication-theorem assumptions.

Batch review, attribution and evidence. Thank you for the contribution.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants